Chapter 5.1 - Lecture Flashcards
What is Artificial Intelligence (AI)?
Artificial Intelligence is technology that allows a machine to mimic human behaviour
What are the subsets of Artificial Intelligence (AI)?
Machine Learning (Supervised) & Deep Learning (Unsupervised)
What is Machine Learning?
Machine Learning is a type of supervised learning, where humans curate training data, and write rules to create algorithms to allow this process to work
What is Deep Learning?
Deep Learning is a type of unsupervised learning that is mediated through an artificial neural network where there is no human intervention
Machine Learning is ___
Supervised
Deep Learning is ___
Unsupervised
___ has enabled faster progress in artificial vision
Deep Learning (Unsupervised)
Humans don’t know enough about ___ to make good rules about machine learning for artificial vision, ___ is able to create a better performing AI
Biological Vision // Unsupervised Deep Learning
What is a consequence of Deep Learning?
Humans sometimes can’t understand the rules of artificial neural networks
How many neurons does a fruit fly have?
~250,000
Which is more advanced, navigation and vision from Computer AI, or from a biological system?
Biological system
How long will it take before our artificial computer-based vision and navigation matches that of a biological system?
A long time
What is the Inverse Projection Problem?
The stimulus falling on the retina is very ambiguous, and different objects can cast the same image on the retina
We exist and operate in a __ world, however we gather visual information from __ images projected onto the retina, and so there can be many possible shapes in the environment that can create a similar image
3D // 2D
Projection of the __ world onto the __ retina, is then reconstructed into a __ Perception
3D // 2D // 3D
The retinal image only provides limited information about the ___ of the object
Properties
Why can the retinal image be difficult to interpret?
Because objects of interest can be hidden/obscured or blurred
What happens if you’re not accommodating to the distance of an object?
Any of the other objects in the scene, closer or further, will be blurred
What is Viewpoint Invariance?
The ability to recognize a single object seen from different viewpoints
What is Viewpoint Invariance important for?
For interacting with objects that can move relative to where you are looking at them from.
Computers used to have tremendous difficulty with ___ , but they are getting better at it
viewpoint invariance
What can human perception effortlessly achieve that computers still struggle with?
Viewpoint Invariance, (recognition of objects, for e.g. faces)
Computers facial recognition software can recognize human faces facing ___ as well as humans can, but do much worse with human faces facing ___
head-on // different directions
What can cause ambiguity in an image?
Light & Shadow
What is Perceptual Organization?
A process where elements in the environment are organized (put together / combined / grouped) to create our perception of objects
What are the two components of Perceptual Organization?
Grouping & Segregation
What is Grouping?
Bringing elements together, putting them together into objects
What is Segregation?
Separating those objects from different parts of the scene
Who was Wilhelm Wundt?
A Structuralist
How did Structuralists refer to Sensations?
Elementary processes that occur from sensory stimulation
How did Structuralists refer to Perception?
Complex conscious experience produced, including awareness of objects
How did Structuralists believe perceptions were created?
By combining individual sensations
Who said that “The whole is the sum of its parts”?
Structuralists
Who said that “The whole is different than the sum of the part”?
Kohler & Wertheimer
What is an example of “The whole is different than the sum of the part”?
Our perception of the 3D world when we know that the images on the retina are actually 2 dimensional
What is Apparent Movement?
A percept that is different than the sum of its parts, because there is no movement that is present, it is just creating the perception of movement.
Illusory Contours and Necker Cubers demonstrate ___
That the whole is different than the sum of its parts
Who came up with the Laws of Perceptual Organization?
The Gestaltists
What is Pragnanz – Law of Good Figure / Law of Simplicity?
Every stimulus is perceived so that the structure is as simple as possible (K.I.S.S - Keep It Simple Stupid)
What is the Law of similarity?
Similar things appear to be grouped together
Law of Good Continuation?
Lines and patterns tend to be perceived as continuing in time and space. Straight or smoothly curving lines are seen as belonging together
(Example: this is a way in which people are able to untangle their headphone cords)
What is Proximity of Nearness?
Things that are near to each other appear to be grouped together.
When a percept creates a battle between two Gestalt laws, will Proximity or Similarity win?
This usually depends on the observer
What 3 laws came after the original Gestalt work?
Common Region, Connectedness, and Synchrony
What is Common Region?
Elements within the same region appear grouped together, uniform
What is Connectedness?
Connected regions are perceived as a single unit
What is Synchrony?
Events that occur at the same time are perceived as belonging together
What is Law of Common Fate?
Objects moving in the same direction tend to be grouped together. Similar to synchrony, but requires motion
What is Perceptual Segregation/Figure-ground Segregation?
Separating objects from the rest of the scene
What is Figure?
A separate object, that stands out from the background
What is Ground?
The background on which the object sits
Figure is perceived as more ___, and is more ___ than the ground
“thing-like” // memorable
Figure is usually seen as being ___ the ground/background
in front of
Ground is perceived as ___, and extends ___ the figure
“unformed” // behind
What is Border Ownership?
The border that separates the figure from the ground is perceived as belonging to the figure
Are symmetrical areas seen as Figure or Ground?
Figure
Are convex things seen as Figure or Ground?
Figure
Are small things seen as Figure or Ground?
Figure
Do things on the bottom of the scene tend to be seen as Figure or Ground?
Figure
Do things on the bottom of the scene tend to be seen as Figure or Ground?
Figure
Why do things on the bottom of the scene tend to be seen as Figure?
Because in natural scenes, most of the objects that you can interact with are on the bottom half of the scene
Gestaltist downplayed the importance of ___
Top-Down/learned Information/experience/meaningfullness
There has been some inconsistency in the way Gestaltists accounted for ___ in Figure-Ground Segregation
meaningfulness
___ can affect figure-ground segregation even for short viewing durations
Learning/Top-Down Information/Meaning
What is Prolonged Viewing?
Familiar objects that we are accustomed to seeing, tend to be grouped and segregated from the ground (sometimes known as Gestalt law of familiarity), meaningfulness is important here
It is useful to view the Gestalt Laws as ___
Heuristics
What is a Heuristic?
A problem solving short-cut, or “best-guess” solution. Heuristics save time and processing power, but only work sometimes
What is an Algorithm?
A set of (sometimes many) steps for solving a problem that works every time. Can be very time and energy intensive